ECJ-INTEREST-L Archives

September 2006

ECJ-INTEREST-L@LISTSERV.GMU.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Sean Luke <[log in to unmask]>
Reply To:
ECJ Evolutionary Computation Toolkit <[log in to unmask]>
Date:
Fri, 29 Sep 2006 01:26:29 -0400
Content-Type:
text/plain
Parts/Attachments:
text/plain (45 lines)
Can you provide a specific set of parameters such that this is  
repeatable in a stock ECJ arrangement?

I suspect you're doing something which is forcing crossover to toss  
its second child.  There are two possibilities here that I can think  
of immediately:
	- you've got "toss" set to true (see Crossover.java)
	- you're feeding Crossover into another pipeline which is only  
asking for one child at a time.  This will force Crossover to throw  
away its second child because the parent pipeline demands only a  
single child.  If you want the second child to be cached and sent  
next time a request is made, you can do this by using BufferedPipeline.

Now as to a new generation that's *smaller* than the expected size --  
this is very highly unlikely unless you've made some serious  
customizations to your parameters.  What if your selection  
procedure?  Your evolution mechanism?

Sean

On Sep 28, 2006, at 10:56 PM, Loretta Macklem wrote:

> Hi,
>
> I am working on a project, where I am trying to track every step  
> where an individual gets added to the new generation. I am using  
> the basic Koza parameters specified in the Koza param file, and for  
> some reason, I am discovering that during crossover, the second new  
> individual is being thrown away. I am then seeing a new population  
> who is not as big as it should be. Does anyone know why this is  
> occurring? Is there something else going on that is filling the  
> spots, and I am just not finding the code that is doing it? I am  
> keeping track of the reproduction individuals, and that is acting  
> normally, but the number of ones produced is not making up for the  
> number of second new individuals being thrown out. I am basically  
> seeing a new generation that is smaller than the number I set in my  
> parameter file. Can anyone offer some help on this? I would greatly  
> appreciate it.
>
> Thank you,
>
> Loretta Macklem
>
> Michigan State University

ATOM RSS1 RSS2